Moving mean in matrix

Hi MATLAB community,

I am using the following code to identify number of 999999, replace to mean of column,

% Identify the columns that contain at least one 999999
isKey = matriz_media == key; 
colIdx = any(isKey,1); 
% Count the number of rows per column that are not 999999
rowCount = sum(~isKey); 
% Temporarily replace 999999 with 0 and calculate the column means
matrixTemp = matriz_media .* ~isKey; 
colMean = sum(matrixTemp)./rowCount;  
colMean=transpose(colMean);
[rowIdx,colIdx] = find(matrixTemp==0); 
matrixTemp(sub2ind(size(matrixTemp),rowIdx,colIdx)) = colMean(colIdx); 

But, I need replace to moving mean of element previous and posterior, such as:

The element of column 5 and line 298 is 999999, I need replace this value to moving mean between (column 5 and element 297 and 299).

Thank you very much!

Guilherme

Answer 1

[dpb]

% Identify the columns that contain at least one 999999
isKey = matriz_media == key; 
colIdx = find(any(isKey,1));            % turn into the column IDs from logical vector
% Replace key value with mean each side of missing value
for c=colIdx
  ix=find(isKey(:,c));                                                   % missing rows
  matriz_media(ix,c)=interp1(matriz_media(:,c),matriz_media(ix,c));      % fill in with linear interpolation
end

Alternatively, w/ R2016b and later...

matriz_media(ismissing(matriz_media,key))=nan;       % create missing values
matriz_media=fillmissing(matriz_media,'movmean',2);  % fill in 2-pt movinvg mean (same as linear interp)

ERRATUM/ADDENDUM:

Oh, yeah...one oversight -- the column vector (:,c) contains all the values including the NaN and interp1 will end up with the same index...you've got to build an x vector without those elements to fill in with the wanted...let's see...

key=999999;
isKey = matriz_media == key; 
colIdx = find(any(isKey,1));            % turn into the column IDs from logical vector
nRow=size(matriz_mdia,1);               % number rows in array
% Replace key value with mean each side of missing value
for c=colIdx
  ix=find(isKey(:,c));                                                   % missing rows
  x=1:nRow;                                                              % all x
  x=x(~ix);                                                              % don't include missing ones
  matriz_media(ix,c)=interp1(x,matriz_media(:,c),matriz_media(ix,c));      % fill in with linear interpolation
end
Z = zscore(matriz_media);

ALERT: Aircode; untested written in forum--check well. :)

The difference is using the x argument as well as y in interp1 which is all indices in array excepting the missing values to interpolate.

[Guilherme Lopes de Campos]

Hi dpb,

Thank for help,

But, not works,

key=999999;
isKey = matriz_media == key; 
colIdx = find(any(isKey,1));            % turn into the column IDs from logical vector
% Replace key value with mean each side of missing value
for c=colIdx
  ix=find(isKey(:,c));                                                   % missing rows
  matriz_media(ix,c)=interp1(matriz_media(:,c),matriz_media(ix,c));      % fill in with linear interpolation
end
Z = zscore(matriz_media);

Those value stayed of (NaN),

Could help me please?

Guilherme

[dpb]

See ADDENDUM to Answer -- altho why not use the ismissing/fillmissing pair (unless you are on an earlier release, maybe?)?